[llvm-commits] [llvm] r51134 - in /llvm/trunk: lib/Transforms/Scalar/InstructionCombining.cpp test/Transforms/InstCombine/2003-11-13-ConstExprCastCall.ll test/Transforms/InstCombine/fp-ret-bitcast.ll

[Duncan Sands]

Hi Chris,

> >>> in your example you can see the function definition (and not just
> >>> a declaration), so doing the transform only when you have the
> >>> declaration of the callee to hand would be ok for this example.
> >>
> >> Ah right, good call.  For declarations, I think the only cases that
> >> matter are things like simplifylibcalls, when the libcall didn't have
> >> the right prototype.  I'm not sure how much this matters.
> >
> > maybe those would be cleaned by my globalopt transform :)
> 
> Why put this in globalopt?  The form that changes the call is is a  
> local/peephole xform, and instcombine is run much more often.
> 
> >>> PS: I just chucked some code into GlobalOpt that teaches it to  
> >>> change
> >>> the type of a declaration if the only use is a bitcast to a function
> >>> type.  It seems to work!  This transforms the objc testcase to:
> >>
> >> The problem is that this objc function can return float, double, and
> >> long double depend on the callee :)
> >
> > If it is called in several different ways in the same module then  
> > indeed
> > there's not much you can do.  Still, I can't help feeling this  
> > transform
> > is useful since I can imagine that quite often people have code in  
> > which
> > some bogus declaration is used but each call calls it in the  
> > "correct" way.
> 
> I don't see the compelling reason to do this.  What use case does it  
> help?

the main reason was that it amused me to write this!  You are right
that it is essentially useless: while it does fire during the gcc
build, it does so very very rarely.  How about I change the instcombine
transform to only change calls when the declaration of the callee is
visible, and leave it at that?

Ciao,

Duncan.